Flow-based motion planning blueprint for autonomous vehicles

Abstract

A system includes a memory device, and a processing device, operatively coupled to the memory device, to receive a set of input data including a representation of a drivable space for an autonomous vehicle (AV), generate, based on the representation of the drivable space, a motion planning blueprint from a flow field modeling the drivable space, and identify, using the motion planning blueprint, a driving path of the AV within the drivable space.

What is claimed is: 
     
       1. A system comprising:
 a memory device; and 
 a processing device, operatively coupled to the memory device, to:
 obtain a representation of a drivable space for an autonomous vehicle (AV); 
 obtain, based on the representation of the drivable space, a set of motion parameters and a set of equipotential contour curves; 
 generate, based on the set of motion parameters and the set of equipotential contour curves, a motion planning blueprint from a flow field modeling the drivable space; 
 identify, using the motion planning blueprint, a driving path of the AV within the drivable space; and 
 cause the AV to follow the driving path. 
 
 
     
     
       2. The system of  claim 1 , wherein:
 the representation of the drivable space comprises a discretized representation of the drivable space comprising a first end corresponding to a traffic inlet and a second end corresponding to a traffic outlet; and 
 to generate the motion planning blueprint, the processing device is further to:
 convert the discretized representation of the drivable space into a flow field data structure; 
 obtain the set of motion parameters and the set of equipotential contour curves based on the flow field data structure; 
 generate a set of geometric samples using the set of motion parameters and the set of equipotential contour curves; and 
 construct the motion planning blueprint based on the set of geometric samples. 
 
 
     
     
       3. The system of  claim 2 , wherein, to obtain the representation of the drivable space, the processing device is further to:
 receive a second representation of the drivable space; and 
 convert the second representation of the drivable space into the discretized representation of the drivable space. 
 
     
     
       4. The system of  claim 2 , wherein:
 to convert the discretized representation into the flow field data structure, the processing device is further to determine a value assigned to each cell of a plurality of cells of the discretized representation; 
 the plurality of cells comprises a first cell having a first value located at the first end, a second cell having a second value located at the second end, and a third cell having a third value located at an intermediate region between the first end and the second end; 
 the second value is greater than the first value; and 
 the third value is between the first value and the second value. 
 
     
     
       5. The system of  claim 1 , wherein the set of motion parameters comprises an interpolated heading of a location within the representation of the drivable space and an interpolated curvature of the location within the representation of the drivable space. 
     
     
       6. The system of  claim 5 , wherein the interpolated heading is obtained as an interpolated gradient corresponding to the location, and wherein the interpolated curvature is obtained from an interpolated Hessian corresponding to the location. 
     
     
       7. The system of  claim 2 , wherein the motion planning blueprint comprises a graph including a plurality of nodes and a plurality of edges, wherein each node of the plurality of nodes corresponds to a geometric sample of the set of geometric samples positioned with respect to a location on a corresponding equipotential contour curve of the set of equipotential contour curves, and wherein the plurality of edges comprises an edge connecting a first node on a first equipotential contour curve reachable to a second node on a second equipotential contour curve different from the first equipotential contour curve. 
     
     
       8. A method comprising:
 obtaining, by a processing device, a representation of a drivable space for an autonomous vehicle (AV); 
 obtaining, by the processing device based on the representation of the drivable space, a set of motion parameters and a set of equipotential contour curves; 
 generating, by the processing device based on the set of motion parameters and the set of equipotential contour curves, a motion planning blueprint from a flow field modeling the drivable space; 
 identifying, by the processing device using the motion planning blueprint, a driving path of the AV within the drivable space; and 
 causing, by the processing device, the AV to follow the driving path. 
 
     
     
       9. The method of  claim 8 , wherein:
 the representation of the drivable space is a discretized representation of the drivable space comprising a first end corresponding to a traffic inlet and a second end corresponding to a traffic outlet; and 
 the generating the motion planning blueprint further comprises:
 converting, by the processing device, the discretized representation of the drivable space into a flow field data structure; 
 obtaining the set of motion parameters and the set of equipotential contour curves based on the discretized representation and the flow field data structure; 
 generating, by the processing device, a set of geometric samples using the set of motion parameters and the set of equipotential contour curves; and 
 constructing the motion planning blueprint based on the set of geometric samples. 
 
 
     
     
       10. The method of  claim 9 , wherein obtaining the representation of the drivable space further comprises:
 receiving, by the processing device, a second representation of the drivable space; and 
 converting the second representation of the drivable space into the discretized representation of the drivable space. 
 
     
     
       11. The method of  claim 9 , wherein:
 converting the discretized representation into the flow field data structure further comprises determining a value assigned to each cell of a plurality of cells of the discretized representation; 
 the plurality of cells comprises a first cell having a first value located at the first end, a second cell having a second value located at the second end, and a third cell having a third value located at an intermediate region between the first end and the second end; 
 the second value is greater than the first value; and 
 the third value is between the first value and the second value. 
 
     
     
       12. The method of  claim 9 , wherein:
 the set of motion parameters comprises an interpolated heading of a location within the discretized representation and an interpolated curvature of the location within the discretized representation; 
 the interpolated heading is obtained as an interpolated gradient corresponding to the location; and 
 the interpolated curvature is obtained from an interpolated Hessian corresponding to the location. 
 
     
     
       13. The method of  claim 9 , wherein the motion planning blueprint comprises a graph including a plurality of nodes and a plurality of edges, wherein each node of the plurality of nodes corresponds to a geometric sample of the set of geometric samples positioned with respect to a location on a corresponding equipotential contour curve of the set of equipotential contour curves, and wherein the plurality of edges comprises an edge connecting a first node on a first equipotential contour curve reachable to a second node on a second equipotential contour curve different from the first equipotential contour curve. 
     
     
       14. A non-transitory computer-readable storage medium having instructions stored thereon that, when executed by a processing device, cause the processing device to perform operations comprising:
 obtaining a discretized representation of a drivable space for an autonomous vehicle (AV), wherein the discretized representation of the drivable space comprises a plurality of cells, a first end corresponding to a traffic inlet and a second end corresponding to a traffic outlet; 
 obtaining, based on the discretized representation of the drivable space, a set of motion parameters and a set of equipotential contour curves, wherein obtaining the set of motion parameters and the set of equipotential contour curves comprises converting the discretized representation of the drivable space into a flow field data structure modeling the drivable space by determining a value assigned to each cell of the plurality of cells of the discretized representation of the drivable space, wherein the first end comprises a first cell having a first value corresponding to a lower bound of a range of values, and wherein the second end comprises a second cell having a second potential value corresponding to an upper bound of the range of values; 
 generating, based on the set of motion parameters and the set of equipotential contour curves, a motion planning blueprint from the flow field data structure; 
 identifying, using the motion planning blueprint, a driving path of the AV within the drivable space; and 
 causing the AV to follow the driving path. 
 
     
     
       15. The non-transitory computer-readable storage medium of  claim 14 , wherein generating the motion planning blueprint further comprises:
 generating a set of geometric samples using the set of interpolated motion parameters and the set of contour curves; and 
 constructing the motion planning blueprint based on the set of geometric samples. 
 
     
     
       16. The non-transitory computer-readable storage medium of  claim 14 , wherein:
 the set of motion parameters comprises an interpolated heading of a location within the discretized representation and an interpolated curvature of the location within the discretized representation; 
 the interpolated heading is obtained as an interpolated gradient corresponding to the location; and 
 the interpolated curvature is obtained from an interpolated Hessian corresponding to the location. 
 
     
     
       17. The non-transitory computer-readable storage medium of  claim 15 , wherein the motion planning blueprint comprises a graph including a plurality of nodes and a plurality of edges, wherein each node of the plurality of nodes corresponds to a geometric sample of the set of geometric samples positioned with respect to a location on a corresponding equipotential contour curve of the set of equipotential contour curves, and wherein the plurality of edges comprises an edge connecting a first node on a first equipotential contour curve reachable to a second node on a second equipotential contour curve different from the first equipotential contour curve. 
     
     
       18. The non-transitory computer-readable storage medium of  claim 14 , wherein obtaining the discretized representation of the drivable space further comprises:
 receiving an input representation of the drivable space; and 
 converting the input representation of the drivable space into the discretized representation of the drivable space.